By Richard M. H. Suen
http://d.repec.org/n?u=RePEc:ucr:wpaper:201004&r=dge
This paper presents a dynamic competitive equilibrium model with heterogeneous time preferences that can account for the observed patterns of wealth and income inequality in the United States. This model generalizes the standard neoclassical growth model by including (i) a demand for status by the consumers and (ii) human capital formation. The first feature prevents the wealth distribution from collapsing into a degenerate distribution. The second feature generates a strong positive correlation between earnings and wealth across agents. A calibrated version of this model succeeds in replicating the wealth and income distributions of the United States.
It is surprisingly difficult to replicate the distribution of wealth. One way to do it is by assuming heterogeneous preferences, but this requires more heterogeneity than what would be reasonable. Richard Suen gets here increasing returns to heterogeneity by adding human capital formation (richer people can get even richer) and wealth in the utility function (people want to hold wealth, not just consume it). Are these reasonable assumptions?
NEP-DGE Blog 
>> “Are these reasonable assumptions?”
Well, no, but what would be an alternative hypothesis and what sort of observations would allow us to discriminate between them?
Accumulation from transfers in random collisions, similar to distribution of heat in a gas, actually doesn’t do so badly, with the right parameters. Everyone flipping coins, with wealth accumulated by the serial “winners” isn’t half bad.
I know that’s not the context in which you want to generate wealth accumulation, but just suggesting it isn’t all that complicated, at base.
Chapter 13 of Cockshott, Cottrell, Michaelson, Wright, and Yakovenko’s Classical Econophysics (Routledge, 2009) explains the distribution of income and wealth to some extent. They have social classes and are interested in statistical equilibrium, as in thermodynamics. I don’t know why one should care about what can do in the failed neoclassical paradigm.
I am really puzzled that this post attracted econophysicians. As I recently highlighted on my blog, they mostly advocated statistical model that are data mining of the worst kind. While you may find some relationships, this does not allow to understand where they come from and especially how they would change if some exogenous variable changes. We do not want to be subject to the Lucas Critique again.
The work he mentions does have models in it, they are indeed relatively parsimonious, which in most sciences is taken as a virtue, and they give good statistical predictions of things like wealth distribution, income distribution, and the rate of return on capital.
Well, the question is badly worded.
It’s easy to replicate the distribution of wealth, as a statistical artifact.
It’s hard to replicate the distribution of wealth in a way that’s meaningful, i.e. in terms of exogenous variation only. That is, produce a model that is not subject to e.g . dumb arbitrage and which has something to say re: policy and welfare.
Could you clarify what you mean when you say that a replication of the distribution of wealth shoudl be meaningful in terms f exogenous variation only. Exogenous variation of what?
[Sorry for the delayed reply.]
I mean that for an explanation to be good it must replicate the distribution by appealing only to variation we can trust to be exogenous: tastes/preferences, technology, endowments and (maybe!) some fundamental institutions.
There should also be purposeful behavior, i.e. lack of any glaring arbitrage opportunities.
Under those restriction, I think that the task of replicating the wealth distribution is hard: you have to assume either a lot of variation in time preference (impatient poor?), large productivity differences (really productive rich?) or some weird market organization (large rents/market imperfections?).